Useful AI and How to Make it Happen

Dayle Hall:  

Hi, and welcome to our podcast, Automating the Enterprise. I’m your host, Dayle Hall, the CMO of SnapLogic. Our podcast is designed to give the organizations out there the insights and best practices on how to integrate, automate and transform their enterprise. Our guest today is a thought leader in tech innovations, particularly in that hot area of artificial intelligence. He believes that AI is a powerful transformation tool for all enterprises, but when used in a very disciplined approach. Currently, he has a team accelerating AI adoption for companies looking at optimizing energy usage and improving their sustainability efforts. Very interesting aspect there, something we haven’t had on this program before. 

So please welcome to the podcast, Sreedhar Sistu, the VP of AI offers at Schneider Electric. Sreedhar, welcome to the podcast.

Sreedhar Sistu: 

Thank you, Dayle. Happy to be here.

Dayle Hall: 

Yeah, we’re happy to have you. I’m going to say this upfront because I’m very proud of the fact that Schneider is actually a customer of ours. So I’m just going to put that out there. But this is not a sales pitch. So anyone that just heard that on this podcast and you’re like, oh, I’ll turn this off, no, no, stick with us, I promise, I promise has nothing to do with buying more SnapLogic. I can see Sreedhar’s face on the video. He’s like, whew, thank goodness. Anyway, we’re excited to have you on here, mate. 

So before we dig into this area, specifically, give me a little bit of background review in the industry, how you actually got into this specific area within Schneider, because this is not something that I would immediately associate with your company.

Sreedhar Sistu:

Yeah, sure. I’ll give from two directions, personally, my journey and then how at Schneider, we arrived at this. I’ve been in the tech industry for quite a long time in enterprise software, cloud SaaS applications, slowly trying to look at the benefits of having SaaS, which is you have data that you can help your customers. However, my experience with AI goes way back, I think for people who are veterans in the industry may remember Lisp and Prolog, which happened to be one of the first few programming languages I learned all the way back in college. What I see is a big transformation happening in software where you used to have applications that are more about the system of record where you store things in one place, and everyone can access it, great value. And then having the right kind of workflow, moving things from one person to another person, one department to another. 

And I think what we are seeing is the next evolution in that, where the data that is generated by people working together is, in fact, going to generate how the workflows and it provide insights for people to become more effective. And that is how I see AS as becoming the next transformative technology. And at Schneider, which- just to give a quick background, Schneider Electric is one of the large energy management companies in the world, we don’t produce electricity, we don’t transmit, but we do everything to distribute energy to buildings, data centers, houses and everything, and manage them much more efficiently. We sell lots of products that are good, I’m sure you see. And we also sell a lot of software. And we have a big IoT platform because all the devices that we sell now are connected devices and they emit data, they have plenty of sensors. And what we see is, hey, using all of this data, we can generate lots of insights. But that’s not something humans can do. This is clearly where AI plays a role. 

So I’ll pause here, and maybe let you delve in a little bit deeper.

Dayle Hall: 

Yeah. No, I think that’s the key part here, which is whether you’re a hardware or a software company, these days, the way devices are connected, the way our applications are connected, it is about data, every company- you’ve heard the expression, every company is a data company. And I think Schneider, specifically, that is very true. Now, obviously, because you’re a customer of SnapLogic, a lot of the work that we do is around- on integration of those data points and so on. 

I think the first question from me would be as you’ve been involved in things like the connection of devices or software applications and so on, how have you seen that, let’s call it, the integration and automation space evolve and change, and how is AI actually becoming a bigger part of that?

Sreedhar Sistu:

Yeah. So I think to your point about application of AI, we can certainly look at what we do in the integration space as a prime candidate for application of AI because this is kind of the next evolution of any kind of enterprise integrations that you do which are putting various pieces together, writing some rules, writing some logic. And guess what? You can probably derive some of that using the data, using how people actually do the integration, which systems they are collecting data from and what they are integrating. 

Now, moving to the other side, which is on the IoT and the connected devices side, where we see a big opportunity from Schneider Electric’s standpoint is, look, we have people doing AI in spaces such as recommendations, selling advertisements and whatnot, it’s a great business model, no questions about that. But where we see a bigger impact is in the space of energy efficiency and automation, right? Today, we all know what is happening in the industry with the energy crisis in Europe. And if we can use technology to improve the utilization of energy, one of the big problems is really peak loads. How do you manage peak loads on the grid so that we don’t shut them down? So there are lots of use cases that we see from our customer standpoint that we think we can address quite effectively using AI, and that’s how we really started the journey on AI.

Dayle Hall:

Yeah. Yeah, if I look at what SnapLogic does, our AI helps to, let’s say, pre-write the connection point so people don’t have to code. They have suggested, if you have these applications, this is how it connects easier. So it’s a use case, it’s an efficiency model. And I like what Schneider is doing, which is they’re looking at not just like, how can we use AI everywhere, it’s like, what are the use cases we can solve for. And that when you talk about things like the energy crisis, that’s a massive global impact. So if you can help to solve these problems on a smaller scale with specific customers and certain regions, that can help these bigger crises that you refer to upfront. So it’s- there’s a little bit of, I don’t want to say it’s an altruistic play too, but there’s definitely- there’s benefit for your company, but there’s also potentially bigger benefit for everyone that’s consuming any type of energy.

Sreedhar Sistu:

You got that right. And that’s one of the things that we are quite proud of in that it is not only about doing things that help us, customers directly in using our asset, but it’s really contributing to your major goal. See, today, the energy demand is only going to increase with increasing population. And then if you look at electrification with more and more electric vehicles coming on board, the demand on the grid is going to grow even higher. And the only way to do that is to make the energy consumption smarter, and that requires technologies such as AI. 

Dayle Hall:

Right, right. You mentioned something earlier when we were talking about integration, automation and using AI, you mentioned specifically business processes. So let’s start with where to start. So think about the most basic aspects of including AI in an enterprise and what should the enterprise really think about when considering integrating any kind of AI into their business processes today? Where do they start? 

Sreedhar Sistu:

Yeah. So there’s a bit of irony with respect to AI today in that it probably is not going to do everything that you thought it will do. 

Dayle Hall:

Really?

Sreedhar Sistu:

Exactly. It will do dishes, it will do cooking. 

Dayle Hall:

Will it take my dog for a walk?

Sreedhar Sistu:

Unfortunately, well, you’ll have to wait for that. You talk to Sony. But it will probably do a lot more than what you think it can, it is capable of. So it’s really about setting realistic expectations on what AI can do. It’s easy to get taken over by these ideas of super intelligence, artificial general intelligence, taking away humans, everything automatically. I think these are all interesting concepts. I am sure we will see them at some point in the future. But today, it is really about doing really cognitively demanding tasks that people are involved, that have to spend a lot of time doing them, and how can we automate them through application of technology. So having that realistic expectation is a key point of adopting in the enterprise. And the second thing that we can delve into deeper later, which is having the right process and methods in place to deploy AI and that includes several things, including identifying the right use cases. So-

Dayle Hall: 

That I think is key. And I think that’s one of the areas I’d like to ask you about specifically. So what I’ve heard from doing these podcasts from talking to customers is value of AI really hits home or has a bigger impact when you start with a use case. So I’m interested to know, as you talk to customers, as you think about how Schneider offers this in the market, how do you advise enterprises to start with the use cases, and do they understand that, and does that make them more successful?

Sreedhar Sistu:  

Yeah. One of the underappreciated things is how hard it is to find real good use cases for AI. Most companies get excited about, hey, we’ve got this new technology, let’s hire a few data scientists and let’s say a few ML engineers, and then we’ll figure something out, right? And you come up with a great demo, you come up with a great proof of concept that everyone likes, but then no one knows how to take it to production, no one knows how to actually use it. And it stays on the demo center for a long time before the next shiny thing comes up. So what we tried to do is, we started taking an approach of, let’s start with use cases, and let’s start with use cases that have a business buy-in. So it’s not a technology push. We are happy to have all the nice technology talk. But let’s start with the business use cases. Let’s make the hypothesis around the business value of these use cases right at the beginning, right? Now, there is a bit of a conundrum there. Some people are afraid that if you have to find business value right upfront, it’s going to be complicated, people are always uncomfortable trying to put values upfront. Now, what we did is, let’s take this next stage-gate approach. So it’s not just saying that, okay, you said something, and we’re going to stay with this forever, but let us start with the initial hypothesis, let us do our first set of exploration to see does it make sense, does it have business value, do we have technical feasibility, and then go into next stage, which gives confidence, which allows us to be a little more, what can I say, open and experimental in approach rather than committing everything forever.

Dayle Hall:

Right. I like that concept. Let me just ask for clarification on that. Because you said, project some things that have business buy-in, but then other one- but then you also said, projects that add- where AI can add value to the business, are those two separate things, are they mutually exclusive? How have you seen that in your own customers?

Sreedhar Sistu:

Yeah. So it’s not so much that they are mutually exclusive. See, what happens is, the one who is on the business side, who is looking for, how do I enhance what we are doing today, right? Let us take a case where we are- we sell software that is trying to optimize energy usage in a building, for example, right? So you can say that there are various ways of optimizing, I can build new features, I can try to find new controls, etc. Or we can say that let us try to find something that is intelligent, that’s going to take leaps and bounds, or the person from the business side may not be fully conversant with the art of the possible with AI. So you might still be looking at more on the incremental side, and this is where we bring in what we call a hub-and-spoke model. So where you have people from the AI side with a technology background and the people in the business side sit together to come up with a proposal where with the AI technologists can explain how AI can help, hypothetically, I mean, we are not yet in the development phase, and someone from business can suddenly see the potential that it can offer to the customers. So this kind of- we have another term called a power couple, so this is how these two can come together to discuss and identify the use cases that makes sense.

Dayle Hall: 

I like that concept to the power couple there, yeah, no, that’s good. And then if we look at- so when you’re out there, when you talk to these examples, how often do you have to potentially go in and help pull those use cases out, or does the- is the business approaching you saying, what, we have them and we want to see whether you can help on that? Is it kind of 50/50 and does it really matter as long as you’re either delivering value on a current project or delivering value to the business?

Sreedhar Sistu:

Yeah. I think you definitely bring a lot of the pragmatist view, Dayle, here because-

Dayle Hall:

I try, I try. 

Sreedhar Sistu:

In a perfect world, every AI engineer or a data scientist would like someone to come with a problem that is well-defined, we call them as Kaggle problems, right? You just bring everything, and then all I have to do is to build the algorithm. 

Dayle Hall:

Yeah, yeah, obviously. 

Sreedhar Sistu:

No, it’s- I think it’s mostly- I think it’s the third option that you said, which it doesn’t really matter who brings kind of the idea and the use case, what is really important is that we work together as one team, you bring the business perspective and you bring the technology side, so we can be sure that, a, it has business value, it is business viable, b, it is technically feasible.

Dayle Hall: 

Right. Yeah. No, I like that. So if I think what we’re talking about here is those kinds of organizations that see the value or are willing to work with it, they’re probably a little bit ahead of the curve. Why do you think in some enterprises or just in general, or do you think we’re behind on the maturity model for AI and putting that into practice in enterprises do you think we’re behind? Or is there a slowness to adopt it, is there still a reticence because they don’t know how to use it? From your experience, where are we on the AI maturity model? And I know there’s many different use cases, so saying it’s all AI is a little bit broad. But what are you seeing?

Sreedhar Sistu:

Yeah. So my experience is, and this is not to make it more self-serving, but I think at an enterprise level, the adoption of AI is generally a little bit lagging, okay? And the reason is several falls, but I can probably postulate a few. One that I mentioned earlier, which is, there’s sometimes a bit of unrealistic expectations what AI can do because we are reading the press, you talk about now- ChatGPT is the next big thing, what ChatGPT is capable of doing certain things and if you were to- some of the output from ChatGPT, sometimes it is confidently wrong.

Dayle Hall:

Yeah. Confidently wrong, that’s a great way of describing it.

Sreedhar Sistu:

So it gives the illusion that the AI behind is really smart when it is not, but it doesn’t know. So I think in the enterprise, what is happening is there is an interest in harness this technology and there is also a bit of a hesitation in adoption. And I can try a contrast with what we have done at Schneider is it is not that Schneider has not been dabbling into AI before. So it’s been around for about five years, we’ve been trying various things. What we started doing was, you can do things at small scale that solve some point solutions, some small problems that still have business value but unless you take a holistic approach and have a comprehensive strategy, you’re not going to scale. And that’s what we decided. And luckily, we have got a strong support from our ExCom that said that you know what, let us take a comprehensive approach, let’s have the right level of investment, let’s have the right operating model to deliver AI at scale. Not a lot of enterprises I think are ready to make that level of commitment. And to me, that’s one of the main reasons why we don’t see as much adoption as we need to.

Dayle Hall:

It’s interesting. When we were doing discussions ahead of time, ahead of this podcast, you said an interesting comment to me, which was having a disciplined approach to implementation. So you talked a little bit there about how organizations should think about it. When you say when- if someone’s out there thinking about, okay, well, we’re looking at AI, maybe they have a use case, or maybe they have the project that they’re looking at. When you describe implementation of this as a disciplined approach, what do you mean by a disciplined approach? 

Sreedhar Sistu:

Yeah. So from my perspective, when I say that we need a disciplined approach for AI at scale is- so we have this an internal process that we call a funnel process, okay? It’s a classic stage gate. I don’t claim that there is something super unique about it. But what is important is to apply that rigorously for everything that goes through. And it is very tempting to look at a new use case comes, oh, we already have the solution for this, there is a third-party solution available, we can implement it right away, or the data scientist is very clever, they can cook up something over the week, boom, done. No, I think we have to go through the stage gate to make sure that everyone is involved, all the stakeholders are involved, and we can move it at the pace that is required to make sure all the elements of it are covered. 

There was a classic paper from Google a couple of years ago that talked about the scope of AI development where we really focus on the data science algorithm development, which is really like 5% to 10% of the work, right? And in enterprise, if you look at it, AI is not a stand-alone product. AI almost always tends to be integrated into something else. And having the how are you going to deliver AI capabilities to end users unless you think about it ahead of time, it won’t happen in the end, or you will be delayed. So when I talk about the disciplined approach, this is to make sure that all bases are covered. And the second aspect of the discipline is really to reject ideas, to kill ideas.

Dayle Hall: 

Explain what you mean by that, kill or reject ideas.

Sreedhar Sistu:

So what is most important is that when we are identifying use cases, you have to be a little bit bold on finding things that have some level of success, not necessarily that it is a slam dunk, because you never know. I mean, if you’re really being honest to yourself until it is done, it’s not done, right? So you pick the ideas. And you know that when you go through a stage, say, an exploration, you run through some sample datasets, you know, you look at business. And you realize that maybe you know what, it’s not so great, or it’s really difficult technically, then let’s drop the idea. And people have this really- feel discomfort in rejecting and dropping ideas because it’s almost seen as a failure. To me, that’s one of the big hindrances. If you’re really disciplined, drop the idea, great exploration done, learn something, not a great business case, let’s stop it.

Dayle Hall:  

Yeah. I think people definitely find it hard to say no, but I think it’s more than that. As they embark on a project, and I think this is where the organization just has to have that discipline or the governance in place that says, yeah, we don’t think we’re going to- whilst AI could help, we don’t think it will have as much value as we think, at least now. But let’s be honest about that and potentially look at other opportunities. But everyone wants their projects and their initiatives to work. So I can imagine that’s interesting. A couple of areas before we finish with our- specifically with Schneider on the sustainability piece. A couple of things that I hear a lot about from these podcasts is one is AI ethics, and how you’re thinking about that specifically. And again, the AI ethics discussions that I had on these podcasts have been very much around the people, at least from an HR perspective and so on, do you have anything to- how does Schneider think about AI ethics with anything related to what you’re offering today?

Sreedhar Sistu:

Yeah. So I mean, it is a topic that we take super seriously. And in fact, right at the formation of central AI organization, we have a group that is really working with the standard bodies and trying to understand the implications of, first of all, to have explainable AI, I think what we really strive towards is, before we can get into ethics, let’s understand what the machine is telling us, right, are super important for us, so we have actually people focusing on explainable AI and making sure that we understand what these algorithms are telling us. Secondly, we do take the data privacy regulations quite seriously. I mean, we have European heritage and GDPR, etc., is really the core of what we do, you’re dealing with sensitive employee data, we have stricter data guidelines in how to anonymize the data for training, etc. So very at the heart of what we do.

Dayle Hall: 

Okay. Yeah, well, that’s good. I think this is an area whilst you hear more about AI in general, I think the AI ethical discussion is just starting to get a little bit more traction. And the thing that I’ve heard, I’m sure you would agree is, the most important thing is that explainability of how this works and not just, hey, we have this black box of AI, and this is how it works, and this is the output. Because then how do you know if there’s bias in there, how do you know if there’s- if it’s actually the right data in, that’s actually the AI-er is making that assessment on? So I think that is- we’re going to see more and more discussions on those. Knowing the way tech companies pop up, I’m sure there’s going to be a bunch of tech companies focused on AI ethics as well at some point in the future, but- 

Sreedhar Sistu:

Oh, yeah. 

Dayle Hall:

[inaudible] Schneider thinks about that, too.

Sreedhar Sistu:

Yeah, yeah, yeah. It’s an [EDI], in fact, we’ve thought right at the beginning, and we started seriously talking about AI that we have to be ahead of the curve on this, and we’ll keep at it to make sure that we are addressing the concerns. 

Dayle Hall: 

Yeah. No, that’s good. Okay. So as we finish, as we wrap up the podcast, let’s talk a little bit specifically about Schneider, your mission with sustainability in AI. So as you’ve already explained, to me, Schneider’s mission is to be the digital partner of enterprises when it comes to sustainability and efficiency, which, by the way, I think is excellent. So what are these projects look like and what are the key things that Schneider is doing to work with enterprises on some specific initiatives like this?

Sreedhar Sistu:

So I can give a couple of examples that actually give some real concreteness to this, right? I think we talked earlier about the challenges that we are facing in the energy space and how we should be able to address. So one of the solutions that we enhance with AI is really around what we call as energy flexibility, okay? So if you think about today, microgrids is an emerging idea for most businesses, buildings, factories, industries, etc. So at the heart of it, it is very simple, right? I mean, you have multiple forms of energy that you can have, distributed energy resources as we call them, because you can have a solar panel that is producing electricity during daytime, you might have a battery that stores it, and you may have a generator for emergencies. You obviously have the electricity coming from the grid. It’s not- what we do is, we can optimize the energy usage from various resources to meet your goals of whether you want to reduce the cost. 

For example, if you have variable charges during daytime, which is the case I’m sure in California, you can optimize when to charge the battery, when to keep the energy with you so that you don’t hit the peak on the grid, you can save using PV production at some parts of the day. And more importantly, with increasing electric vehicles in the buildings, you can also charge them more optimally, not every vehicle needs to be charged in half an hour, right? It’s going to be in the parking lot for all day, so maybe you can drip charge it. So one of the solutions that we provide is to take all these factors into account and forecast the demand using machine learning algorithms and then really try to reduce the peaks in the energy consumption. And either you can reduce the electricity bill that you pay, or you can also favor a lot more of the sustainable energy so that you can reduce the CO2 footprint. 

Dayle Hall:

Interesting. 

Sreedhar Sistu:

So I think this is a really powerful solution that both helps you to optimize the cost as well as reach your sustainability goals.

Dayle Hall:

Yeah. I like that. I’m seeing more and more the companies springing up around sustainability and there’s obviously, particularly in Europe, which is where Schneider is founded and headquartered, massive opportunity as well as rules and regulations coming down. So I think you’re definitely ahead of the curve. How much would you say- no, I don’t want you to- you don’t have to tell me Schneider’s corporate strategy as it were. But is this something that’s definitely- it doesn’t sound like it’s a side project in Schneider. It sounds like it’s going to become a leading initiative. How much effort is Schneider putting into this type of area?

Sreedhar Sistu:

So we have a dedicated line of business that is focusing on microgrids. We also have a significant focus on at [FCs]. Recently, we launched a new home energy management system. So we are even going from the commercial buildings into residential places where you can have an optimal energy usage within the homes. So we have a big strategy around what we call as prosumer because see, homes no longer are just consumers of electricity from the grid. Homes can produce electricity using PV. So that changes the game completely. And we think we have a strong role to play in that transformation, which is both an opportunity for us as a business as well as a healthy environment.

Dayle Hall:

Yeah. No, I love that. It’s a great- and look, we’re always- we’re happy that you’re a SnapLogic customer, but I love that thinking from a massive energy customer that they’re thinking about the sustainability and supporting the average household even if they’re not buying Schneider necessarily directly as a product or service. So look, that’s excellent. I have one final question for you as we wrap up. So you’ve been in and around this industry, I put it- I’d describe you as a leader in tech innovations, particularly in AI. So as a leader in tech innovations, particularly in AI, what are you excited about not just with Schneider, but in general, the opportunities around AI in the next, I don’t know, two to five years, what are you looking forward to seeing, what are you excited about within your area of expertise?

Sreedhar Sistu:

First of all, I am super bullish on the benefits of AI in coming years. And I see really the shift towards, what I call it, a useful AI rather than the tight AI that we had in the last maybe 510 years, which was necessary to bring maturity into technology, but I think we start to see more and more useful applications of AI. And really, there, it’s about augmenting humans, it’s a call it, a, it’s not about replacing humans but augmenting them. There’s certain things that humans are good at and will continue to be good at, and AI is going to help. Do you want someone to be watching on the shop floor looking for defects with their eyes? No. You want a camera that has embedded AI to detect defects and push items on the site. So I think we are going to start seeing a lot of useful AI applications and that’s going to enhance productivity efficiency and, as I said, towards, in our case, sustainability of the planet, so I’m super bullish about it.

Dayle Hall:  

I think that is a fitting end to the podcast. I love what you said, useful AI rather than a toy. The most important thing, which is, AI will help to augment humans, not remove them from the process and give us- as humans, we’ll be able to focus on more interesting opportunities, the more interesting works. That’s a perfect way to end. Sreedhar, it’s been my pleasure. And hopefully, it hasn’t been too painful to walk through what your focus is around AI, what Schneider are doing. Thank you so much for being part of our podcast today.

Sreedhar Sistu:

Thank you, Dayle. It was a great conversation. I enjoyed every bit of it and appreciate having me on your podcast.

Dayle Hall: 

Sounds good. To everyone else out there, thanks for joining us today. Check us out on our next podcast that will be coming up soon. And have a good rest of your day.